Pileus Git - ~andy/linux/blob - fs/ext2/super.c

   1 /*
   2  *  linux/fs/ext2/super.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  from
  10  *
  11  *  linux/fs/minix/inode.c
  12  *
  13  *  Copyright (C) 1991, 1992  Linus Torvalds
  14  *
  15  *  Big-endian to little-endian byte-swapping/bitmaps by
  16  *        David S. Miller (davem@caip.rutgers.edu), 1995
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/string.h>
  21 #include <linux/fs.h>
  22 #include <linux/slab.h>
  23 #include <linux/init.h>
  24 #include <linux/blkdev.h>
  25 #include <linux/parser.h>
  26 #include <linux/random.h>
  27 #include <linux/buffer_head.h>
  28 #include <linux/exportfs.h>
  29 #include <linux/vfs.h>
  30 #include <linux/seq_file.h>
  31 #include <linux/mount.h>
  32 #include <linux/log2.h>
  33 #include <linux/quotaops.h>
  34 #include <asm/uaccess.h>
  35 #include "ext2.h"
  36 #include "xattr.h"
  37 #include "acl.h"
  38 #include "xip.h"
  39
  40 static void ext2_sync_super(struct super_block *sb,
  41                             struct ext2_super_block *es, int wait);
  42 static int ext2_remount (struct super_block * sb, int * flags, char * data);
  43 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  44 static int ext2_sync_fs(struct super_block *sb, int wait);
  45 static int ext2_freeze(struct super_block *sb);
  46 static int ext2_unfreeze(struct super_block *sb);
  47
  48 void ext2_error(struct super_block *sb, const char *function,
  49                 const char *fmt, ...)
  50 {
  51         struct va_format vaf;
  52         va_list args;
  53         struct ext2_sb_info *sbi = EXT2_SB(sb);
  54         struct ext2_super_block *es = sbi->s_es;
  55
  56         if (!(sb->s_flags & MS_RDONLY)) {
  57                 spin_lock(&sbi->s_lock);
  58                 sbi->s_mount_state |= EXT2_ERROR_FS;
  59                 es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  60                 spin_unlock(&sbi->s_lock);
  61                 ext2_sync_super(sb, es, 1);
  62         }
  63
  64         va_start(args, fmt);
  65
  66         vaf.fmt = fmt;
  67         vaf.va = &args;
  68
  69         printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  70                sb->s_id, function, &vaf);
  71
  72         va_end(args);
  73
  74         if (test_opt(sb, ERRORS_PANIC))
  75                 panic("EXT2-fs: panic from previous error\n");
  76         if (test_opt(sb, ERRORS_RO)) {
  77                 ext2_msg(sb, KERN_CRIT,
  78                              "error: remounting filesystem read-only");
  79                 sb->s_flags |= MS_RDONLY;
  80         }
  81 }
  82
  83 void ext2_msg(struct super_block *sb, const char *prefix,
  84                 const char *fmt, ...)
  85 {
  86         struct va_format vaf;
  87         va_list args;
  88
  89         va_start(args, fmt);
  90
  91         vaf.fmt = fmt;
  92         vaf.va = &args;
  93
  94         printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  95
  96         va_end(args);
  97 }
  98
  99 /*
 100  * This must be called with sbi->s_lock held.
 101  */
 102 void ext2_update_dynamic_rev(struct super_block *sb)
 103 {
 104         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
 105
 106         if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
 107                 return;
 108
 109         ext2_msg(sb, KERN_WARNING,
 110                      "warning: updating to rev %d because of "
 111                      "new feature flag, running e2fsck is recommended",
 112                      EXT2_DYNAMIC_REV);
 113
 114         es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
 115         es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
 116         es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
 117         /* leave es->s_feature_*compat flags alone */
 118         /* es->s_uuid will be set by e2fsck if empty */
 119
 120         /*
 121          * The rest of the superblock fields should be zero, and if not it
 122          * means they are likely already in use, so leave them alone.  We
 123          * can leave it up to e2fsck to clean up any inconsistencies there.
 124          */
 125 }
 126
 127 static void ext2_put_super (struct super_block * sb)
 128 {
 129         int db_count;
 130         int i;
 131         struct ext2_sb_info *sbi = EXT2_SB(sb);
 132
 133         dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 134
 135         ext2_xattr_put_super(sb);
 136         if (!(sb->s_flags & MS_RDONLY)) {
 137                 struct ext2_super_block *es = sbi->s_es;
 138
 139                 spin_lock(&sbi->s_lock);
 140                 es->s_state = cpu_to_le16(sbi->s_mount_state);
 141                 spin_unlock(&sbi->s_lock);
 142                 ext2_sync_super(sb, es, 1);
 143         }
 144         db_count = sbi->s_gdb_count;
 145         for (i = 0; i < db_count; i++)
 146                 if (sbi->s_group_desc[i])
 147                         brelse (sbi->s_group_desc[i]);
 148         kfree(sbi->s_group_desc);
 149         kfree(sbi->s_debts);
 150         percpu_counter_destroy(&sbi->s_freeblocks_counter);
 151         percpu_counter_destroy(&sbi->s_freeinodes_counter);
 152         percpu_counter_destroy(&sbi->s_dirs_counter);
 153         brelse (sbi->s_sbh);
 154         sb->s_fs_info = NULL;
 155         kfree(sbi->s_blockgroup_lock);
 156         kfree(sbi);
 157 }
 158
 159 static struct kmem_cache * ext2_inode_cachep;
 160
 161 static struct inode *ext2_alloc_inode(struct super_block *sb)
 162 {
 163         struct ext2_inode_info *ei;
 164         ei = (struct ext2_inode_info *)kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
 165         if (!ei)
 166                 return NULL;
 167         ei->i_block_alloc_info = NULL;
 168         ei->vfs_inode.i_version = 1;
 169         return &ei->vfs_inode;
 170 }
 171
 172 static void ext2_i_callback(struct rcu_head *head)
 173 {
 174         struct inode *inode = container_of(head, struct inode, i_rcu);
 175         kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
 176 }
 177
 178 static void ext2_destroy_inode(struct inode *inode)
 179 {
 180         call_rcu(&inode->i_rcu, ext2_i_callback);
 181 }
 182
 183 static void init_once(void *foo)
 184 {
 185         struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
 186
 187         rwlock_init(&ei->i_meta_lock);
 188 #ifdef CONFIG_EXT2_FS_XATTR
 189         init_rwsem(&ei->xattr_sem);
 190 #endif
 191         mutex_init(&ei->truncate_mutex);
 192         inode_init_once(&ei->vfs_inode);
 193 }
 194
 195 static int init_inodecache(void)
 196 {
 197         ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
 198                                              sizeof(struct ext2_inode_info),
 199                                              0, (SLAB_RECLAIM_ACCOUNT|
 200                                                 SLAB_MEM_SPREAD),
 201                                              init_once);
 202         if (ext2_inode_cachep == NULL)
 203                 return -ENOMEM;
 204         return 0;
 205 }
 206
 207 static void destroy_inodecache(void)
 208 {
 209         kmem_cache_destroy(ext2_inode_cachep);
 210 }
 211
 212 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
 213 {
 214         struct super_block *sb = root->d_sb;
 215         struct ext2_sb_info *sbi = EXT2_SB(sb);
 216         struct ext2_super_block *es = sbi->s_es;
 217         unsigned long def_mount_opts;
 218
 219         spin_lock(&sbi->s_lock);
 220         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 221
 222         if (sbi->s_sb_block != 1)
 223                 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
 224         if (test_opt(sb, MINIX_DF))
 225                 seq_puts(seq, ",minixdf");
 226         if (test_opt(sb, GRPID))
 227                 seq_puts(seq, ",grpid");
 228         if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
 229                 seq_puts(seq, ",nogrpid");
 230         if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
 231             le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
 232                 seq_printf(seq, ",resuid=%u",
 233                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
 234         }
 235         if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
 236             le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
 237                 seq_printf(seq, ",resgid=%u",
 238                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
 239         }
 240         if (test_opt(sb, ERRORS_RO)) {
 241                 int def_errors = le16_to_cpu(es->s_errors);
 242
 243                 if (def_errors == EXT2_ERRORS_PANIC ||
 244                     def_errors == EXT2_ERRORS_CONTINUE) {
 245                         seq_puts(seq, ",errors=remount-ro");
 246                 }
 247         }
 248         if (test_opt(sb, ERRORS_CONT))
 249                 seq_puts(seq, ",errors=continue");
 250         if (test_opt(sb, ERRORS_PANIC))
 251                 seq_puts(seq, ",errors=panic");
 252         if (test_opt(sb, NO_UID32))
 253                 seq_puts(seq, ",nouid32");
 254         if (test_opt(sb, DEBUG))
 255                 seq_puts(seq, ",debug");
 256         if (test_opt(sb, OLDALLOC))
 257                 seq_puts(seq, ",oldalloc");
 258
 259 #ifdef CONFIG_EXT2_FS_XATTR
 260         if (test_opt(sb, XATTR_USER))
 261                 seq_puts(seq, ",user_xattr");
 262         if (!test_opt(sb, XATTR_USER) &&
 263             (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
 264                 seq_puts(seq, ",nouser_xattr");
 265         }
 266 #endif
 267
 268 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 269         if (test_opt(sb, POSIX_ACL))
 270                 seq_puts(seq, ",acl");
 271         if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
 272                 seq_puts(seq, ",noacl");
 273 #endif
 274
 275         if (test_opt(sb, NOBH))
 276                 seq_puts(seq, ",nobh");
 277
 278 #if defined(CONFIG_QUOTA)
 279         if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
 280                 seq_puts(seq, ",usrquota");
 281
 282         if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
 283                 seq_puts(seq, ",grpquota");
 284 #endif
 285
 286 #if defined(CONFIG_EXT2_FS_XIP)
 287         if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
 288                 seq_puts(seq, ",xip");
 289 #endif
 290
 291         if (!test_opt(sb, RESERVATION))
 292                 seq_puts(seq, ",noreservation");
 293
 294         spin_unlock(&sbi->s_lock);
 295         return 0;
 296 }
 297
 298 #ifdef CONFIG_QUOTA
 299 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
 300 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
 301 #endif
 302
 303 static const struct super_operations ext2_sops = {
 304         .alloc_inode    = ext2_alloc_inode,
 305         .destroy_inode  = ext2_destroy_inode,
 306         .write_inode    = ext2_write_inode,
 307         .evict_inode    = ext2_evict_inode,
 308         .put_super      = ext2_put_super,
 309         .sync_fs        = ext2_sync_fs,
 310         .freeze_fs      = ext2_freeze,
 311         .unfreeze_fs    = ext2_unfreeze,
 312         .statfs         = ext2_statfs,
 313         .remount_fs     = ext2_remount,
 314         .show_options   = ext2_show_options,
 315 #ifdef CONFIG_QUOTA
 316         .quota_read     = ext2_quota_read,
 317         .quota_write    = ext2_quota_write,
 318 #endif
 319 };
 320
 321 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
 322                 u64 ino, u32 generation)
 323 {
 324         struct inode *inode;
 325
 326         if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
 327                 return ERR_PTR(-ESTALE);
 328         if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
 329                 return ERR_PTR(-ESTALE);
 330
 331         /*
 332          * ext2_iget isn't quite right if the inode is currently unallocated!
 333          * However ext2_iget currently does appropriate checks to handle stale
 334          * inodes so everything is OK.
 335          */
 336         inode = ext2_iget(sb, ino);
 337         if (IS_ERR(inode))
 338                 return ERR_CAST(inode);
 339         if (generation && inode->i_generation != generation) {
 340                 /* we didn't find the right inode.. */
 341                 iput(inode);
 342                 return ERR_PTR(-ESTALE);
 343         }
 344         return inode;
 345 }
 346
 347 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
 348                 int fh_len, int fh_type)
 349 {
 350         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 351                                     ext2_nfs_get_inode);
 352 }
 353
 354 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
 355                 int fh_len, int fh_type)
 356 {
 357         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 358                                     ext2_nfs_get_inode);
 359 }
 360
 361 static const struct export_operations ext2_export_ops = {
 362         .fh_to_dentry = ext2_fh_to_dentry,
 363         .fh_to_parent = ext2_fh_to_parent,
 364         .get_parent = ext2_get_parent,
 365 };
 366
 367 static unsigned long get_sb_block(void **data)
 368 {
 369         unsigned long   sb_block;
 370         char            *options = (char *) *data;
 371
 372         if (!options || strncmp(options, "sb=", 3) != 0)
 373                 return 1;       /* Default location */
 374         options += 3;
 375         sb_block = simple_strtoul(options, &options, 0);
 376         if (*options && *options != ',') {
 377                 printk("EXT2-fs: Invalid sb specification: %s\n",
 378                        (char *) *data);
 379                 return 1;
 380         }
 381         if (*options == ',')
 382                 options++;
 383         *data = (void *) options;
 384         return sb_block;
 385 }
 386
 387 enum {
 388         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
 389         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
 390         Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
 391         Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
 392         Opt_acl, Opt_noacl, Opt_xip, Opt_ignore, Opt_err, Opt_quota,
 393         Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
 394 };
 395
 396 static const match_table_t tokens = {
 397         {Opt_bsd_df, "bsddf"},
 398         {Opt_minix_df, "minixdf"},
 399         {Opt_grpid, "grpid"},
 400         {Opt_grpid, "bsdgroups"},
 401         {Opt_nogrpid, "nogrpid"},
 402         {Opt_nogrpid, "sysvgroups"},
 403         {Opt_resgid, "resgid=%u"},
 404         {Opt_resuid, "resuid=%u"},
 405         {Opt_sb, "sb=%u"},
 406         {Opt_err_cont, "errors=continue"},
 407         {Opt_err_panic, "errors=panic"},
 408         {Opt_err_ro, "errors=remount-ro"},
 409         {Opt_nouid32, "nouid32"},
 410         {Opt_nocheck, "check=none"},
 411         {Opt_nocheck, "nocheck"},
 412         {Opt_debug, "debug"},
 413         {Opt_oldalloc, "oldalloc"},
 414         {Opt_orlov, "orlov"},
 415         {Opt_nobh, "nobh"},
 416         {Opt_user_xattr, "user_xattr"},
 417         {Opt_nouser_xattr, "nouser_xattr"},
 418         {Opt_acl, "acl"},
 419         {Opt_noacl, "noacl"},
 420         {Opt_xip, "xip"},
 421         {Opt_grpquota, "grpquota"},
 422         {Opt_ignore, "noquota"},
 423         {Opt_quota, "quota"},
 424         {Opt_usrquota, "usrquota"},
 425         {Opt_reservation, "reservation"},
 426         {Opt_noreservation, "noreservation"},
 427         {Opt_err, NULL}
 428 };
 429
 430 static int parse_options(char *options, struct super_block *sb)
 431 {
 432         char *p;
 433         struct ext2_sb_info *sbi = EXT2_SB(sb);
 434         substring_t args[MAX_OPT_ARGS];
 435         int option;
 436         kuid_t uid;
 437         kgid_t gid;
 438
 439         if (!options)
 440                 return 1;
 441
 442         while ((p = strsep (&options, ",")) != NULL) {
 443                 int token;
 444                 if (!*p)
 445                         continue;
 446
 447                 token = match_token(p, tokens, args);
 448                 switch (token) {
 449                 case Opt_bsd_df:
 450                         clear_opt (sbi->s_mount_opt, MINIX_DF);
 451                         break;
 452                 case Opt_minix_df:
 453                         set_opt (sbi->s_mount_opt, MINIX_DF);
 454                         break;
 455                 case Opt_grpid:
 456                         set_opt (sbi->s_mount_opt, GRPID);
 457                         break;
 458                 case Opt_nogrpid:
 459                         clear_opt (sbi->s_mount_opt, GRPID);
 460                         break;
 461                 case Opt_resuid:
 462                         if (match_int(&args[0], &option))
 463                                 return 0;
 464                         uid = make_kuid(current_user_ns(), option);
 465                         if (!uid_valid(uid)) {
 466                                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
 467                                 return -1;
 468
 469                         }
 470                         sbi->s_resuid = uid;
 471                         break;
 472                 case Opt_resgid:
 473                         if (match_int(&args[0], &option))
 474                                 return 0;
 475                         gid = make_kgid(current_user_ns(), option);
 476                         if (!gid_valid(gid)) {
 477                                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
 478                                 return -1;
 479                         }
 480                         sbi->s_resgid = gid;
 481                         break;
 482                 case Opt_sb:
 483                         /* handled by get_sb_block() instead of here */
 484                         /* *sb_block = match_int(&args[0]); */
 485                         break;
 486                 case Opt_err_panic:
 487                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 488                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 489                         set_opt (sbi->s_mount_opt, ERRORS_PANIC);
 490                         break;
 491                 case Opt_err_ro:
 492                         clear_opt (sbi->s_mount_opt, ERRORS_CONT);
 493                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 494                         set_opt (sbi->s_mount_opt, ERRORS_RO);
 495                         break;
 496                 case Opt_err_cont:
 497                         clear_opt (sbi->s_mount_opt, ERRORS_RO);
 498                         clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
 499                         set_opt (sbi->s_mount_opt, ERRORS_CONT);
 500                         break;
 501                 case Opt_nouid32:
 502                         set_opt (sbi->s_mount_opt, NO_UID32);
 503                         break;
 504                 case Opt_nocheck:
 505                         clear_opt (sbi->s_mount_opt, CHECK);
 506                         break;
 507                 case Opt_debug:
 508                         set_opt (sbi->s_mount_opt, DEBUG);
 509                         break;
 510                 case Opt_oldalloc:
 511                         set_opt (sbi->s_mount_opt, OLDALLOC);
 512                         break;
 513                 case Opt_orlov:
 514                         clear_opt (sbi->s_mount_opt, OLDALLOC);
 515                         break;
 516                 case Opt_nobh:
 517                         set_opt (sbi->s_mount_opt, NOBH);
 518                         break;
 519 #ifdef CONFIG_EXT2_FS_XATTR
 520                 case Opt_user_xattr:
 521                         set_opt (sbi->s_mount_opt, XATTR_USER);
 522                         break;
 523                 case Opt_nouser_xattr:
 524                         clear_opt (sbi->s_mount_opt, XATTR_USER);
 525                         break;
 526 #else
 527                 case Opt_user_xattr:
 528                 case Opt_nouser_xattr:
 529                         ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
 530                                 "not supported");
 531                         break;
 532 #endif
 533 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 534                 case Opt_acl:
 535                         set_opt(sbi->s_mount_opt, POSIX_ACL);
 536                         break;
 537                 case Opt_noacl:
 538                         clear_opt(sbi->s_mount_opt, POSIX_ACL);
 539                         break;
 540 #else
 541                 case Opt_acl:
 542                 case Opt_noacl:
 543                         ext2_msg(sb, KERN_INFO,
 544                                 "(no)acl options not supported");
 545                         break;
 546 #endif
 547                 case Opt_xip:
 548 #ifdef CONFIG_EXT2_FS_XIP
 549                         set_opt (sbi->s_mount_opt, XIP);
 550 #else
 551                         ext2_msg(sb, KERN_INFO, "xip option not supported");
 552 #endif
 553                         break;
 554
 555 #if defined(CONFIG_QUOTA)
 556                 case Opt_quota:
 557                 case Opt_usrquota:
 558                         set_opt(sbi->s_mount_opt, USRQUOTA);
 559                         break;
 560
 561                 case Opt_grpquota:
 562                         set_opt(sbi->s_mount_opt, GRPQUOTA);
 563                         break;
 564 #else
 565                 case Opt_quota:
 566                 case Opt_usrquota:
 567                 case Opt_grpquota:
 568                         ext2_msg(sb, KERN_INFO,
 569                                 "quota operations not supported");
 570                         break;
 571 #endif
 572
 573                 case Opt_reservation:
 574                         set_opt(sbi->s_mount_opt, RESERVATION);
 575                         ext2_msg(sb, KERN_INFO, "reservations ON");
 576                         break;
 577                 case Opt_noreservation:
 578                         clear_opt(sbi->s_mount_opt, RESERVATION);
 579                         ext2_msg(sb, KERN_INFO, "reservations OFF");
 580                         break;
 581                 case Opt_ignore:
 582                         break;
 583                 default:
 584                         return 0;
 585                 }
 586         }
 587         return 1;
 588 }
 589
 590 static int ext2_setup_super (struct super_block * sb,
 591                               struct ext2_super_block * es,
 592                               int read_only)
 593 {
 594         int res = 0;
 595         struct ext2_sb_info *sbi = EXT2_SB(sb);
 596
 597         if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
 598                 ext2_msg(sb, KERN_ERR,
 599                         "error: revision level too high, "
 600                         "forcing read-only mode");
 601                 res = MS_RDONLY;
 602         }
 603         if (read_only)
 604                 return res;
 605         if (!(sbi->s_mount_state & EXT2_VALID_FS))
 606                 ext2_msg(sb, KERN_WARNING,
 607                         "warning: mounting unchecked fs, "
 608                         "running e2fsck is recommended");
 609         else if ((sbi->s_mount_state & EXT2_ERROR_FS))
 610                 ext2_msg(sb, KERN_WARNING,
 611                         "warning: mounting fs with errors, "
 612                         "running e2fsck is recommended");
 613         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
 614                  le16_to_cpu(es->s_mnt_count) >=
 615                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
 616                 ext2_msg(sb, KERN_WARNING,
 617                         "warning: maximal mount count reached, "
 618                         "running e2fsck is recommended");
 619         else if (le32_to_cpu(es->s_checkinterval) &&
 620                 (le32_to_cpu(es->s_lastcheck) +
 621                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
 622                 ext2_msg(sb, KERN_WARNING,
 623                         "warning: checktime reached, "
 624                         "running e2fsck is recommended");
 625         if (!le16_to_cpu(es->s_max_mnt_count))
 626                 es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
 627         le16_add_cpu(&es->s_mnt_count, 1);
 628         if (test_opt (sb, DEBUG))
 629                 ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
 630                         "bpg=%lu, ipg=%lu, mo=%04lx]",
 631                         EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
 632                         sbi->s_frag_size,
 633                         sbi->s_groups_count,
 634                         EXT2_BLOCKS_PER_GROUP(sb),
 635                         EXT2_INODES_PER_GROUP(sb),
 636                         sbi->s_mount_opt);
 637         return res;
 638 }
 639
 640 static int ext2_check_descriptors(struct super_block *sb)
 641 {
 642         int i;
 643         struct ext2_sb_info *sbi = EXT2_SB(sb);
 644
 645         ext2_debug ("Checking group descriptors");
 646
 647         for (i = 0; i < sbi->s_groups_count; i++) {
 648                 struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
 649                 ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
 650                 ext2_fsblk_t last_block;
 651
 652                 if (i == sbi->s_groups_count - 1)
 653                         last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
 654                 else
 655                         last_block = first_block +
 656                                 (EXT2_BLOCKS_PER_GROUP(sb) - 1);
 657
 658                 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
 659                     le32_to_cpu(gdp->bg_block_bitmap) > last_block)
 660                 {
 661                         ext2_error (sb, "ext2_check_descriptors",
 662                                     "Block bitmap for group %d"
 663                                     " not in group (block %lu)!",
 664                                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
 665                         return 0;
 666                 }
 667                 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
 668                     le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
 669                 {
 670                         ext2_error (sb, "ext2_check_descriptors",
 671                                     "Inode bitmap for group %d"
 672                                     " not in group (block %lu)!",
 673                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
 674                         return 0;
 675                 }
 676                 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
 677                     le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
 678                     last_block)
 679                 {
 680                         ext2_error (sb, "ext2_check_descriptors",
 681                                     "Inode table for group %d"
 682                                     " not in group (block %lu)!",
 683                                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
 684                         return 0;
 685                 }
 686         }
 687         return 1;
 688 }
 689
 690 /*
 691  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
 692  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
 693  * We need to be 1 filesystem block less than the 2^32 sector limit.
 694  */
 695 static loff_t ext2_max_size(int bits)
 696 {
 697         loff_t res = EXT2_NDIR_BLOCKS;
 698         int meta_blocks;
 699         loff_t upper_limit;
 700
 701         /* This is calculated to be the largest file size for a
 702          * dense, file such that the total number of
 703          * sectors in the file, including data and all indirect blocks,
 704          * does not exceed 2^32 -1
 705          * __u32 i_blocks representing the total number of
 706          * 512 bytes blocks of the file
 707          */
 708         upper_limit = (1LL << 32) - 1;
 709
 710         /* total blocks in file system block size */
 711         upper_limit >>= (bits - 9);
 712
 713
 714         /* indirect blocks */
 715         meta_blocks = 1;
 716         /* double indirect blocks */
 717         meta_blocks += 1 + (1LL << (bits-2));
 718         /* tripple indirect blocks */
 719         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
 720
 721         upper_limit -= meta_blocks;
 722         upper_limit <<= bits;
 723
 724         res += 1LL << (bits-2);
 725         res += 1LL << (2*(bits-2));
 726         res += 1LL << (3*(bits-2));
 727         res <<= bits;
 728         if (res > upper_limit)
 729                 res = upper_limit;
 730
 731         if (res > MAX_LFS_FILESIZE)
 732                 res = MAX_LFS_FILESIZE;
 733
 734         return res;
 735 }
 736
 737 static unsigned long descriptor_loc(struct super_block *sb,
 738                                     unsigned long logic_sb_block,
 739                                     int nr)
 740 {
 741         struct ext2_sb_info *sbi = EXT2_SB(sb);
 742         unsigned long bg, first_meta_bg;
 743         int has_super = 0;
 744
 745         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
 746
 747         if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
 748             nr < first_meta_bg)
 749                 return (logic_sb_block + nr + 1);
 750         bg = sbi->s_desc_per_block * nr;
 751         if (ext2_bg_has_super(sb, bg))
 752                 has_super = 1;
 753
 754         return ext2_group_first_block_no(sb, bg) + has_super;
 755 }
 756
 757 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 758 {
 759         struct buffer_head * bh;
 760         struct ext2_sb_info * sbi;
 761         struct ext2_super_block * es;
 762         struct inode *root;
 763         unsigned long block;
 764         unsigned long sb_block = get_sb_block(&data);
 765         unsigned long logic_sb_block;
 766         unsigned long offset = 0;
 767         unsigned long def_mount_opts;
 768         long ret = -EINVAL;
 769         int blocksize = BLOCK_SIZE;
 770         int db_count;
 771         int i, j;
 772         __le32 features;
 773         int err;
 774
 775         err = -ENOMEM;
 776         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 777         if (!sbi)
 778                 goto failed_unlock;
 779
 780         sbi->s_blockgroup_lock =
 781                 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
 782         if (!sbi->s_blockgroup_lock) {
 783                 kfree(sbi);
 784                 goto failed_unlock;
 785         }
 786         sb->s_fs_info = sbi;
 787         sbi->s_sb_block = sb_block;
 788
 789         spin_lock_init(&sbi->s_lock);
 790
 791         /*
 792          * See what the current blocksize for the device is, and
 793          * use that as the blocksize.  Otherwise (or if the blocksize
 794          * is smaller than the default) use the default.
 795          * This is important for devices that have a hardware
 796          * sectorsize that is larger than the default.
 797          */
 798         blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
 799         if (!blocksize) {
 800                 ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
 801                 goto failed_sbi;
 802         }
 803
 804         /*
 805          * If the superblock doesn't start on a hardware sector boundary,
 806          * calculate the offset.
 807          */
 808         if (blocksize != BLOCK_SIZE) {
 809                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 810                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 811         } else {
 812                 logic_sb_block = sb_block;
 813         }
 814
 815         if (!(bh = sb_bread(sb, logic_sb_block))) {
 816                 ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
 817                 goto failed_sbi;
 818         }
 819         /*
 820          * Note: s_es must be initialized as soon as possible because
 821          *       some ext2 macro-instructions depend on its value
 822          */
 823         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 824         sbi->s_es = es;
 825         sb->s_magic = le16_to_cpu(es->s_magic);
 826
 827         if (sb->s_magic != EXT2_SUPER_MAGIC)
 828                 goto cantfind_ext2;
 829
 830         /* Set defaults before we parse the mount options */
 831         def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
 832         if (def_mount_opts & EXT2_DEFM_DEBUG)
 833                 set_opt(sbi->s_mount_opt, DEBUG);
 834         if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
 835                 set_opt(sbi->s_mount_opt, GRPID);
 836         if (def_mount_opts & EXT2_DEFM_UID16)
 837                 set_opt(sbi->s_mount_opt, NO_UID32);
 838 #ifdef CONFIG_EXT2_FS_XATTR
 839         if (def_mount_opts & EXT2_DEFM_XATTR_USER)
 840                 set_opt(sbi->s_mount_opt, XATTR_USER);
 841 #endif
 842 #ifdef CONFIG_EXT2_FS_POSIX_ACL
 843         if (def_mount_opts & EXT2_DEFM_ACL)
 844                 set_opt(sbi->s_mount_opt, POSIX_ACL);
 845 #endif
 846
 847         if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
 848                 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
 849         else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
 850                 set_opt(sbi->s_mount_opt, ERRORS_CONT);
 851         else
 852                 set_opt(sbi->s_mount_opt, ERRORS_RO);
 853
 854         sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
 855         sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
 856
 857         set_opt(sbi->s_mount_opt, RESERVATION);
 858
 859         if (!parse_options((char *) data, sb))
 860                 goto failed_mount;
 861
 862         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 863                 ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
 864                  MS_POSIXACL : 0);
 865
 866         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
 867                                     EXT2_MOUNT_XIP if not */
 868
 869         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
 870             (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
 871              EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
 872              EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
 873                 ext2_msg(sb, KERN_WARNING,
 874                         "warning: feature flags set on rev 0 fs, "
 875                         "running e2fsck is recommended");
 876         /*
 877          * Check feature flags regardless of the revision level, since we
 878          * previously didn't change the revision level when setting the flags,
 879          * so there is a chance incompat flags are set on a rev 0 filesystem.
 880          */
 881         features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
 882         if (features) {
 883                 ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
 884                        "unsupported optional features (%x)",
 885                         le32_to_cpu(features));
 886                 goto failed_mount;
 887         }
 888         if (!(sb->s_flags & MS_RDONLY) &&
 889             (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
 890                 ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
 891                        "unsupported optional features (%x)",
 892                        le32_to_cpu(features));
 893                 goto failed_mount;
 894         }
 895
 896         blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
 897
 898         if (ext2_use_xip(sb) && blocksize != PAGE_SIZE) {
 899                 if (!silent)
 900                         ext2_msg(sb, KERN_ERR,
 901                                 "error: unsupported blocksize for xip");
 902                 goto failed_mount;
 903         }
 904
 905         /* If the blocksize doesn't match, re-read the thing.. */
 906         if (sb->s_blocksize != blocksize) {
 907                 brelse(bh);
 908
 909                 if (!sb_set_blocksize(sb, blocksize)) {
 910                         ext2_msg(sb, KERN_ERR,
 911                                 "error: bad blocksize %d", blocksize);
 912                         goto failed_sbi;
 913                 }
 914
 915                 logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
 916                 offset = (sb_block*BLOCK_SIZE) % blocksize;
 917                 bh = sb_bread(sb, logic_sb_block);
 918                 if(!bh) {
 919                         ext2_msg(sb, KERN_ERR, "error: couldn't read"
 920                                 "superblock on 2nd try");
 921                         goto failed_sbi;
 922                 }
 923                 es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
 924                 sbi->s_es = es;
 925                 if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
 926                         ext2_msg(sb, KERN_ERR, "error: magic mismatch");
 927                         goto failed_mount;
 928                 }
 929         }
 930
 931         sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
 932         sb->s_max_links = EXT2_LINK_MAX;
 933
 934         if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
 935                 sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
 936                 sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
 937         } else {
 938                 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
 939                 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
 940                 if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
 941                     !is_power_of_2(sbi->s_inode_size) ||
 942                     (sbi->s_inode_size > blocksize)) {
 943                         ext2_msg(sb, KERN_ERR,
 944                                 "error: unsupported inode size: %d",
 945                                 sbi->s_inode_size);
 946                         goto failed_mount;
 947                 }
 948         }
 949
 950         sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
 951                                    le32_to_cpu(es->s_log_frag_size);
 952         if (sbi->s_frag_size == 0)
 953                 goto cantfind_ext2;
 954         sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
 955
 956         sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
 957         sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
 958         sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
 959
 960         if (EXT2_INODE_SIZE(sb) == 0)
 961                 goto cantfind_ext2;
 962         sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
 963         if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
 964                 goto cantfind_ext2;
 965         sbi->s_itb_per_group = sbi->s_inodes_per_group /
 966                                         sbi->s_inodes_per_block;
 967         sbi->s_desc_per_block = sb->s_blocksize /
 968                                         sizeof (struct ext2_group_desc);
 969         sbi->s_sbh = bh;
 970         sbi->s_mount_state = le16_to_cpu(es->s_state);
 971         sbi->s_addr_per_block_bits =
 972                 ilog2 (EXT2_ADDR_PER_BLOCK(sb));
 973         sbi->s_desc_per_block_bits =
 974                 ilog2 (EXT2_DESC_PER_BLOCK(sb));
 975
 976         if (sb->s_magic != EXT2_SUPER_MAGIC)
 977                 goto cantfind_ext2;
 978
 979         if (sb->s_blocksize != bh->b_size) {
 980                 if (!silent)
 981                         ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
 982                 goto failed_mount;
 983         }
 984
 985         if (sb->s_blocksize != sbi->s_frag_size) {
 986                 ext2_msg(sb, KERN_ERR,
 987                         "error: fragsize %lu != blocksize %lu"
 988                         "(not supported yet)",
 989                         sbi->s_frag_size, sb->s_blocksize);
 990                 goto failed_mount;
 991         }
 992
 993         if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
 994                 ext2_msg(sb, KERN_ERR,
 995                         "error: #blocks per group too big: %lu",
 996                         sbi->s_blocks_per_group);
 997                 goto failed_mount;
 998         }
 999         if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1000                 ext2_msg(sb, KERN_ERR,
1001                         "error: #fragments per group too big: %lu",
1002                         sbi->s_frags_per_group);
1003                 goto failed_mount;
1004         }
1005         if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1006                 ext2_msg(sb, KERN_ERR,
1007                         "error: #inodes per group too big: %lu",
1008                         sbi->s_inodes_per_group);
1009                 goto failed_mount;
1010         }
1011
1012         if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1013                 goto cantfind_ext2;
1014         sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1015                                 le32_to_cpu(es->s_first_data_block) - 1)
1016                                         / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1017         db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1018                    EXT2_DESC_PER_BLOCK(sb);
1019         sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
1020         if (sbi->s_group_desc == NULL) {
1021                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1022                 goto failed_mount;
1023         }
1024         bgl_lock_init(sbi->s_blockgroup_lock);
1025         sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1026         if (!sbi->s_debts) {
1027                 ext2_msg(sb, KERN_ERR, "error: not enough memory");
1028                 goto failed_mount_group_desc;
1029         }
1030         for (i = 0; i < db_count; i++) {
1031                 block = descriptor_loc(sb, logic_sb_block, i);
1032                 sbi->s_group_desc[i] = sb_bread(sb, block);
1033                 if (!sbi->s_group_desc[i]) {
1034                         for (j = 0; j < i; j++)
1035                                 brelse (sbi->s_group_desc[j]);
1036                         ext2_msg(sb, KERN_ERR,
1037                                 "error: unable to read group descriptors");
1038                         goto failed_mount_group_desc;
1039                 }
1040         }
1041         if (!ext2_check_descriptors (sb)) {
1042                 ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1043                 goto failed_mount2;
1044         }
1045         sbi->s_gdb_count = db_count;
1046         get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1047         spin_lock_init(&sbi->s_next_gen_lock);
1048
1049         /* per fileystem reservation list head & lock */
1050         spin_lock_init(&sbi->s_rsv_window_lock);
1051         sbi->s_rsv_window_root = RB_ROOT;
1052         /*
1053          * Add a single, static dummy reservation to the start of the
1054          * reservation window list --- it gives us a placeholder for
1055          * append-at-start-of-list which makes the allocation logic
1056          * _much_ simpler.
1057          */
1058         sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1059         sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1060         sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1061         sbi->s_rsv_window_head.rsv_goal_size = 0;
1062         ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1063
1064         err = percpu_counter_init(&sbi->s_freeblocks_counter,
1065                                 ext2_count_free_blocks(sb));
1066         if (!err) {
1067                 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1068                                 ext2_count_free_inodes(sb));
1069         }
1070         if (!err) {
1071                 err = percpu_counter_init(&sbi->s_dirs_counter,
1072                                 ext2_count_dirs(sb));
1073         }
1074         if (err) {
1075                 ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1076                 goto failed_mount3;
1077         }
1078         /*
1079          * set up enough so that it can read an inode
1080          */
1081         sb->s_op = &ext2_sops;
1082         sb->s_export_op = &ext2_export_ops;
1083         sb->s_xattr = ext2_xattr_handlers;
1084
1085 #ifdef CONFIG_QUOTA
1086         sb->dq_op = &dquot_operations;
1087         sb->s_qcop = &dquot_quotactl_ops;
1088 #endif
1089
1090         root = ext2_iget(sb, EXT2_ROOT_INO);
1091         if (IS_ERR(root)) {
1092                 ret = PTR_ERR(root);
1093                 goto failed_mount3;
1094         }
1095         if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1096                 iput(root);
1097                 ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1098                 goto failed_mount3;
1099         }
1100
1101         sb->s_root = d_make_root(root);
1102         if (!sb->s_root) {
1103                 ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1104                 ret = -ENOMEM;
1105                 goto failed_mount3;
1106         }
1107         if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1108                 ext2_msg(sb, KERN_WARNING,
1109                         "warning: mounting ext3 filesystem as ext2");
1110         if (ext2_setup_super (sb, es, sb->s_flags & MS_RDONLY))
1111                 sb->s_flags |= MS_RDONLY;
1112         ext2_write_super(sb);
1113         return 0;
1114
1115 cantfind_ext2:
1116         if (!silent)
1117                 ext2_msg(sb, KERN_ERR,
1118                         "error: can't find an ext2 filesystem on dev %s.",
1119                         sb->s_id);
1120         goto failed_mount;
1121 failed_mount3:
1122         percpu_counter_destroy(&sbi->s_freeblocks_counter);
1123         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1124         percpu_counter_destroy(&sbi->s_dirs_counter);
1125 failed_mount2:
1126         for (i = 0; i < db_count; i++)
1127                 brelse(sbi->s_group_desc[i]);
1128 failed_mount_group_desc:
1129         kfree(sbi->s_group_desc);
1130         kfree(sbi->s_debts);
1131 failed_mount:
1132         brelse(bh);
1133 failed_sbi:
1134         sb->s_fs_info = NULL;
1135         kfree(sbi->s_blockgroup_lock);
1136         kfree(sbi);
1137 failed_unlock:
1138         return ret;
1139 }
1140
1141 static void ext2_clear_super_error(struct super_block *sb)
1142 {
1143         struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1144
1145         if (buffer_write_io_error(sbh)) {
1146                 /*
1147                  * Oh, dear.  A previous attempt to write the
1148                  * superblock failed.  This could happen because the
1149                  * USB device was yanked out.  Or it could happen to
1150                  * be a transient write error and maybe the block will
1151                  * be remapped.  Nothing we can do but to retry the
1152                  * write and hope for the best.
1153                  */
1154                 ext2_msg(sb, KERN_ERR,
1155                        "previous I/O error to superblock detected\n");
1156                 clear_buffer_write_io_error(sbh);
1157                 set_buffer_uptodate(sbh);
1158         }
1159 }
1160
1161 static void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1162                             int wait)
1163 {
1164         ext2_clear_super_error(sb);
1165         spin_lock(&EXT2_SB(sb)->s_lock);
1166         es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1167         es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1168         es->s_wtime = cpu_to_le32(get_seconds());
1169         /* unlock before we do IO */
1170         spin_unlock(&EXT2_SB(sb)->s_lock);
1171         mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1172         if (wait)
1173                 sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1174 }
1175
1176 /*
1177  * In the second extended file system, it is not necessary to
1178  * write the super block since we use a mapping of the
1179  * disk super block in a buffer.
1180  *
1181  * However, this function is still used to set the fs valid
1182  * flags to 0.  We need to set this flag to 0 since the fs
1183  * may have been checked while mounted and e2fsck may have
1184  * set s_state to EXT2_VALID_FS after some corrections.
1185  */
1186 static int ext2_sync_fs(struct super_block *sb, int wait)
1187 {
1188         struct ext2_sb_info *sbi = EXT2_SB(sb);
1189         struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1190
1191         /*
1192          * Write quota structures to quota file, sync_blockdev() will write
1193          * them to disk later
1194          */
1195         dquot_writeback_dquots(sb, -1);
1196
1197         spin_lock(&sbi->s_lock);
1198         if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1199                 ext2_debug("setting valid to 0\n");
1200                 es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1201         }
1202         spin_unlock(&sbi->s_lock);
1203         ext2_sync_super(sb, es, wait);
1204         return 0;
1205 }
1206
1207 static int ext2_freeze(struct super_block *sb)
1208 {
1209         struct ext2_sb_info *sbi = EXT2_SB(sb);
1210
1211         /*
1212          * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1213          * because we have unattached inodes and thus filesystem is not fully
1214          * consistent.
1215          */
1216         if (atomic_long_read(&sb->s_remove_count)) {
1217                 ext2_sync_fs(sb, 1);
1218                 return 0;
1219         }
1220         /* Set EXT2_FS_VALID flag */
1221         spin_lock(&sbi->s_lock);
1222         sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1223         spin_unlock(&sbi->s_lock);
1224         ext2_sync_super(sb, sbi->s_es, 1);
1225
1226         return 0;
1227 }
1228
1229 static int ext2_unfreeze(struct super_block *sb)
1230 {
1231         /* Just write sb to clear EXT2_VALID_FS flag */
1232         ext2_write_super(sb);
1233
1234         return 0;
1235 }
1236
1237 void ext2_write_super(struct super_block *sb)
1238 {
1239         if (!(sb->s_flags & MS_RDONLY))
1240                 ext2_sync_fs(sb, 1);
1241 }
1242
1243 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1244 {
1245         struct ext2_sb_info * sbi = EXT2_SB(sb);
1246         struct ext2_super_block * es;
1247         unsigned long old_mount_opt = sbi->s_mount_opt;
1248         struct ext2_mount_options old_opts;
1249         unsigned long old_sb_flags;
1250         int err;
1251
1252         spin_lock(&sbi->s_lock);
1253
1254         /* Store the old options */
1255         old_sb_flags = sb->s_flags;
1256         old_opts.s_mount_opt = sbi->s_mount_opt;
1257         old_opts.s_resuid = sbi->s_resuid;
1258         old_opts.s_resgid = sbi->s_resgid;
1259
1260         /*
1261          * Allow the "check" option to be passed as a remount option.
1262          */
1263         if (!parse_options(data, sb)) {
1264                 err = -EINVAL;
1265                 goto restore_opts;
1266         }
1267
1268         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1269                 ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1270
1271         ext2_xip_verify_sb(sb); /* see if bdev supports xip, unset
1272                                     EXT2_MOUNT_XIP if not */
1273
1274         if ((ext2_use_xip(sb)) && (sb->s_blocksize != PAGE_SIZE)) {
1275                 ext2_msg(sb, KERN_WARNING,
1276                         "warning: unsupported blocksize for xip");
1277                 err = -EINVAL;
1278                 goto restore_opts;
1279         }
1280
1281         es = sbi->s_es;
1282         if ((sbi->s_mount_opt ^ old_mount_opt) & EXT2_MOUNT_XIP) {
1283                 ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1284                          "xip flag with busy inodes while remounting");
1285                 sbi->s_mount_opt &= ~EXT2_MOUNT_XIP;
1286                 sbi->s_mount_opt |= old_mount_opt & EXT2_MOUNT_XIP;
1287         }
1288         if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) {
1289                 spin_unlock(&sbi->s_lock);
1290                 return 0;
1291         }
1292         if (*flags & MS_RDONLY) {
1293                 if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1294                     !(sbi->s_mount_state & EXT2_VALID_FS)) {
1295                         spin_unlock(&sbi->s_lock);
1296                         return 0;
1297                 }
1298
1299                 /*
1300                  * OK, we are remounting a valid rw partition rdonly, so set
1301                  * the rdonly flag and then mark the partition as valid again.
1302                  */
1303                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1304                 es->s_mtime = cpu_to_le32(get_seconds());
1305                 spin_unlock(&sbi->s_lock);
1306
1307                 err = dquot_suspend(sb, -1);
1308                 if (err < 0) {
1309                         spin_lock(&sbi->s_lock);
1310                         goto restore_opts;
1311                 }
1312
1313                 ext2_sync_super(sb, es, 1);
1314         } else {
1315                 __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1316                                                ~EXT2_FEATURE_RO_COMPAT_SUPP);
1317                 if (ret) {
1318                         ext2_msg(sb, KERN_WARNING,
1319                                 "warning: couldn't remount RDWR because of "
1320                                 "unsupported optional features (%x).",
1321                                 le32_to_cpu(ret));
1322                         err = -EROFS;
1323                         goto restore_opts;
1324                 }
1325                 /*
1326                  * Mounting a RDONLY partition read-write, so reread and
1327                  * store the current valid flag.  (It may have been changed
1328                  * by e2fsck since we originally mounted the partition.)
1329                  */
1330                 sbi->s_mount_state = le16_to_cpu(es->s_state);
1331                 if (!ext2_setup_super (sb, es, 0))
1332                         sb->s_flags &= ~MS_RDONLY;
1333                 spin_unlock(&sbi->s_lock);
1334
1335                 ext2_write_super(sb);
1336
1337                 dquot_resume(sb, -1);
1338         }
1339
1340         return 0;
1341 restore_opts:
1342         sbi->s_mount_opt = old_opts.s_mount_opt;
1343         sbi->s_resuid = old_opts.s_resuid;
1344         sbi->s_resgid = old_opts.s_resgid;
1345         sb->s_flags = old_sb_flags;
1346         spin_unlock(&sbi->s_lock);
1347         return err;
1348 }
1349
1350 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1351 {
1352         struct super_block *sb = dentry->d_sb;
1353         struct ext2_sb_info *sbi = EXT2_SB(sb);
1354         struct ext2_super_block *es = sbi->s_es;
1355         u64 fsid;
1356
1357         spin_lock(&sbi->s_lock);
1358
1359         if (test_opt (sb, MINIX_DF))
1360                 sbi->s_overhead_last = 0;
1361         else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1362                 unsigned long i, overhead = 0;
1363                 smp_rmb();
1364
1365                 /*
1366                  * Compute the overhead (FS structures). This is constant
1367                  * for a given filesystem unless the number of block groups
1368                  * changes so we cache the previous value until it does.
1369                  */
1370
1371                 /*
1372                  * All of the blocks before first_data_block are
1373                  * overhead
1374                  */
1375                 overhead = le32_to_cpu(es->s_first_data_block);
1376
1377                 /*
1378                  * Add the overhead attributed to the superblock and
1379                  * block group descriptors.  If the sparse superblocks
1380                  * feature is turned on, then not all groups have this.
1381                  */
1382                 for (i = 0; i < sbi->s_groups_count; i++)
1383                         overhead += ext2_bg_has_super(sb, i) +
1384                                 ext2_bg_num_gdb(sb, i);
1385
1386                 /*
1387                  * Every block group has an inode bitmap, a block
1388                  * bitmap, and an inode table.
1389                  */
1390                 overhead += (sbi->s_groups_count *
1391                              (2 + sbi->s_itb_per_group));
1392                 sbi->s_overhead_last = overhead;
1393                 smp_wmb();
1394                 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1395         }
1396
1397         buf->f_type = EXT2_SUPER_MAGIC;
1398         buf->f_bsize = sb->s_blocksize;
1399         buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1400         buf->f_bfree = ext2_count_free_blocks(sb);
1401         es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1402         buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1403         if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1404                 buf->f_bavail = 0;
1405         buf->f_files = le32_to_cpu(es->s_inodes_count);
1406         buf->f_ffree = ext2_count_free_inodes(sb);
1407         es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1408         buf->f_namelen = EXT2_NAME_LEN;
1409         fsid = le64_to_cpup((void *)es->s_uuid) ^
1410                le64_to_cpup((void *)es->s_uuid + sizeof(u64));
1411         buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1412         buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1413         spin_unlock(&sbi->s_lock);
1414         return 0;
1415 }
1416
1417 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1418         int flags, const char *dev_name, void *data)
1419 {
1420         return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1421 }
1422
1423 #ifdef CONFIG_QUOTA
1424
1425 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1426  * acquiring the locks... As quota files are never truncated and quota code
1427  * itself serializes the operations (and no one else should touch the files)
1428  * we don't have to be afraid of races */
1429 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1430                                size_t len, loff_t off)
1431 {
1432         struct inode *inode = sb_dqopt(sb)->files[type];
1433         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1434         int err = 0;
1435         int offset = off & (sb->s_blocksize - 1);
1436         int tocopy;
1437         size_t toread;
1438         struct buffer_head tmp_bh;
1439         struct buffer_head *bh;
1440         loff_t i_size = i_size_read(inode);
1441
1442         if (off > i_size)
1443                 return 0;
1444         if (off+len > i_size)
1445                 len = i_size-off;
1446         toread = len;
1447         while (toread > 0) {
1448                 tocopy = sb->s_blocksize - offset < toread ?
1449                                 sb->s_blocksize - offset : toread;
1450
1451                 tmp_bh.b_state = 0;
1452                 tmp_bh.b_size = sb->s_blocksize;
1453                 err = ext2_get_block(inode, blk, &tmp_bh, 0);
1454                 if (err < 0)
1455                         return err;
1456                 if (!buffer_mapped(&tmp_bh))    /* A hole? */
1457                         memset(data, 0, tocopy);
1458                 else {
1459                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1460                         if (!bh)
1461                                 return -EIO;
1462                         memcpy(data, bh->b_data+offset, tocopy);
1463                         brelse(bh);
1464                 }
1465                 offset = 0;
1466                 toread -= tocopy;
1467                 data += tocopy;
1468                 blk++;
1469         }
1470         return len;
1471 }
1472
1473 /* Write to quotafile */
1474 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1475                                 const char *data, size_t len, loff_t off)
1476 {
1477         struct inode *inode = sb_dqopt(sb)->files[type];
1478         sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1479         int err = 0;
1480         int offset = off & (sb->s_blocksize - 1);
1481         int tocopy;
1482         size_t towrite = len;
1483         struct buffer_head tmp_bh;
1484         struct buffer_head *bh;
1485
1486         while (towrite > 0) {
1487                 tocopy = sb->s_blocksize - offset < towrite ?
1488                                 sb->s_blocksize - offset : towrite;
1489
1490                 tmp_bh.b_state = 0;
1491                 err = ext2_get_block(inode, blk, &tmp_bh, 1);
1492                 if (err < 0)
1493                         goto out;
1494                 if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1495                         bh = sb_bread(sb, tmp_bh.b_blocknr);
1496                 else
1497                         bh = sb_getblk(sb, tmp_bh.b_blocknr);
1498                 if (!bh) {
1499                         err = -EIO;
1500                         goto out;
1501                 }
1502                 lock_buffer(bh);
1503                 memcpy(bh->b_data+offset, data, tocopy);
1504                 flush_dcache_page(bh->b_page);
1505                 set_buffer_uptodate(bh);
1506                 mark_buffer_dirty(bh);
1507                 unlock_buffer(bh);
1508                 brelse(bh);
1509                 offset = 0;
1510                 towrite -= tocopy;
1511                 data += tocopy;
1512                 blk++;
1513         }
1514 out:
1515         if (len == towrite)
1516                 return err;
1517         if (inode->i_size < off+len-towrite)
1518                 i_size_write(inode, off+len-towrite);
1519         inode->i_version++;
1520         inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1521         mark_inode_dirty(inode);
1522         return len - towrite;
1523 }
1524
1525 #endif
1526
1527 static struct file_system_type ext2_fs_type = {
1528         .owner          = THIS_MODULE,
1529         .name           = "ext2",
1530         .mount          = ext2_mount,
1531         .kill_sb        = kill_block_super,
1532         .fs_flags       = FS_REQUIRES_DEV,
1533 };
1534
1535 static int __init init_ext2_fs(void)
1536 {
1537         int err = init_ext2_xattr();
1538         if (err)
1539                 return err;
1540         err = init_inodecache();
1541         if (err)
1542                 goto out1;
1543         err = register_filesystem(&ext2_fs_type);
1544         if (err)
1545                 goto out;
1546         return 0;
1547 out:
1548         destroy_inodecache();
1549 out1:
1550         exit_ext2_xattr();
1551         return err;
1552 }
1553
1554 static void __exit exit_ext2_fs(void)
1555 {
1556         unregister_filesystem(&ext2_fs_type);
1557         destroy_inodecache();
1558         exit_ext2_xattr();
1559 }
1560
1561 MODULE_AUTHOR("Remy Card and others");
1562 MODULE_DESCRIPTION("Second Extended Filesystem");
1563 MODULE_LICENSE("GPL");
1564 module_init(init_ext2_fs)
1565 module_exit(exit_ext2_fs)